Re: How to (of can I) use prediction profiler for setting specification limits and test yield enhancement?This suggestion is not related directly to your question... but if you have access to the raw data, you could check if Interactive Capability Plot would be useful
Names Default To Here(1);
dt = Open("$SAMPLE_DATA/Semiconductor Capability.jmp");
obj = Process Capability(Process Variables(:PNP1), Individual Detail Reports(1));
obj << {:PNP1 << Process Capability Analysis(Interactive Capability Plot(1))};
Also Process Capability platforms Capability Box Plots can be helpful (as can the whole platform Process Capability (jmp.com))
Re: How to (of can I) use prediction profiler for setting specification limits and test yield enhancement?@jthi Thanks. I have access to the raw data but the Cpks have been calculated after filtering the outliers (for each parameter data) using a separate JMP script (no easily accessible to me so do not know what outlier filtering scheme has been used in it).
Is there a simple/recommended outlier filter recipe in JMP which I can use on the raw data before using your suggestion so that I am at least closely aligned to the Cpks already obtained?
Re: How to (of can I) use prediction profiler for setting specification limits and test yield enhancement?You could check out Explore Outliers platform
There is also good blog post series Outliers Episode 4: Detecting outliers using jackknife distance
Re: How to (of can I) use prediction profiler for setting specification limits and test yield enhancement?@jthi The interactive capability appearw to provide ability to tweak process performance Pp, Ppk by changing the limits using the overall sigma. This is useful for longer term process variation analysis. I need to understand how the process capability Cp, Cpk are affected by the spec limits.
The following two appear to be a better route for me unless I have misunderstood something.
- https://community.jmp.com/t5/JMP-Add-Ins/Capability-Analysis-JMP-Add-In/ta-p/22798
- https://community.jmp.com/t5/Discussions/Find-new-spec-limits-by-adjusting-the-CpK-value/m-p/571796/highlight/true#M78139
The fist one does not seem to work for me on JMP 16 and the second one appears to involve manual steps (I will see I I can code this up in JSL).
Is there something in-built in JMP 16 (I am on 16.2) already which allows me to explore Cpk variation by changing LSL or USL or both?
Re: How to (of can I) use prediction profiler for setting specification limits and test yield enhancement?If I understand correctly, you have what you need to calculate Cpk you could create columns for LSL, USL and then formula columns to calculate Cpk? You could also script interactive solution for this, but if you already have the few statistics needed for Cpk you could maybe even use Excel which can be better for simple calculations like this which require user input (unless you script it in JMP).
How to (or can I) use prediction profiler for setting specification limits and test yield enhancement?I have a data set with columns of mean, median, sigma, LSL (lower spec limit), USL (upper spec limit), Cp, Cpk and Parameter Yield Loss [%] for 10 measured parameters for parts (~5000 identically built parts, part IDs are rows). Some parameter Cpks are below 1.33 (may or may not have low Cp). It appears from initial data analysis that the limits are not correctly set.
Using the prediction profiler, I would like to manually nudge the LSL and/or USL to raise the the Cpk to > 1.33 which I am expecting will also reduce the the yield loss for each parameter.
I am very little experience with the prediction profiler so need some guidance on how to use it for my case, if it could be used.
I am thinking of having the LSL and USL knobs (which I can change) on the x-axis of the profiler separately and Cpk of each parameter on the y-axis (and perhaps parameter yield on a secondary y-axis for each parameter).
Advanced Statistical ModelingAutomation and ScriptingBasic Data Analysis and ModelingPredictive Modeling and Machine LearningQuality and Process EngineeringWindows
